Why building the AI version of legacy software isn't enough
November 26, 2025
As a VC you literally can't avoid running into AI startups everywhere, mostly building the "AI version" of some software. These can be either vertical software (law, shipping, finance, compliance, school accreditation, etc.) or horizontal (writing, presenting, researching, etc.) but the pitch is the same:
Take a vertical or horizontal
Point out the market size (wow 100 billion!!!)
Point out how ancient the software is (trash!!!)
So of course we should build an AI-powered version of it.
But IMO that's exactly the wrong way to go about it.
For one, every incumbent is doing precisely that: building an AI version of their existing software, and I hate to break it to you: they can do it just as well as you, AND they have more data, more distribution, and more engineers.
*Yes yes, every startup believes incumbents are just bolting on AI, while they are “true AI natives.” Frankly I don’t know what that means, and in too many cases we’ve seen, it’s just talking points without tangible differentiation to back it up.
Worse, in many cases there are competitors much scarier than the old guard - growth stage startups that have the weight and influence of incumbents, plus the agility and innovation of startups.
So the question is: if you’re a startup, what exactly is your edge?
And to frame the question for us investors, which “AI version of xyz software” should we think about backing?
Here’s my take:
AI is actually the ONE thing you shouldn’t focus on.
Ummm what? Okay, so… what should I focus on?
At this stage of my thinking (and it’s early), I lean towards two things:
  1. Reimagining the workflow—how people actually do the work.
  1. Reimagining the interface—the surface where that work happens.
The bad news is this is a lot harder than just building the AI version of whatever. The good news? If you succeed, it’s much more influential, much harder for incumbents to copy, and much more retentive.
1. Let’s start with workflows
Take Notion for instance
  • It changed how we worked with documents. You don’t shift tab between docs and sheets and the browser, you just press / and bring up the block you need, all on one surface.
  • And you don’t save and mail it to people, you copied the link to share it or publish it.
  • The UX is so influential it became a default for millions of software that followed. Notion wasn’t a feature upgrade. It was a rewiring of how people did knowledge work.
Think of how Figma changed design
  • It wasn’t just “design in the browser.” It made design multiplayer. It turned what used to be a solo-and-handoff process into a real-time team sport.
  • Designers and developers could sit in the same file and iterate live. That one change to the workflow unlocked a completely new speed and energy in product teams.
A more extreme case is Looker (which Google acquired for $2.6 billion in 2019)
  • When Looker first launched, it asked users—especially data analysts—to learn LookML, a whole new modeling language.
  • That’s a big ask. But by learning that new layer, teams unlocked a new way to model data once and use it everywhere. Investors thought they were nuts.
2. What about the interface?
This is the more fluffy and nebulous point compared to workflows, and therefore secondary, right?
Wrong.
I think it could actually be more powerful.
Let me illustrate this with a question I’ve asked way too many Notion users:
“If Microsoft Office added all the same features as Notion, would you adopt it as your main workspace?”
I've asked hundreds. No one has said yes.
And that’s because software is more than a checklist of features.
My feeling is that the feeling of using a piece of software is incredibly important. It shouldn’t be a surprise. You wake up every day and spend upwards 10-20 hours in that virtual space, so how that space makes you feel ought to matter. And while incumbents can copy startups’ features, they can’t copy that feeling.
Hosting your event on Luma feels fundamentally different from doing it on Eventbrite. Setting up Typeform feels fundamentally different from doing so on SurveyMonkey.
At its most extreme, I suspect, your software becomes your identity. I don’t use Microsoft because… well, I’m just not a Microsoft kind of person (yes I have to use Office or Teams occasionally, but trust me I’m never happy about it).
The case of Gamma
I’m writing about this because I'm an investor and a fervent user of Gamma, so no I’m not neutral on this point, but hear me out.
Every few months our team gets a scare because some big company built an AI presentation feature.
OMG Canva slides added AI!
OMG Power Point added AI!
Holy crap Figma (Figma!!!) added slides!
Oh shoot Manus and Perplexity can now generate decks!
And every single time, Gamma grew faster.
*The latest instance is Google Gemini being able to generate slides. Who knows, maybe this time’ll be different!?
So here’s what I think is going on.
I believe Gamma has been resilient because it has changed both the workflow and the interface I prefer. Yes the AI is amazing, and yes AI is what gave Gamma an incredible growth spurt back in 2023, but as a user I almost consider it secondary.
Gamma just looks different. Like Notion, it’s a new kind of thing. It’s slides, docs, websites, all that, and somewhere in-between. More important for me as a user, it feels different. And that’s not something Google Slides or PPT can replicate, no matter how much AI they add.
And that’s also why every month I hear someone say Gamma "doesn't look right to them", and cannot be used to produce professional decks. That just tells me these people want a more traditional look and feel to their decks - e.g. McKinsey strategy slides for clients.
When they say “Gamma isn’t good enough”, they mean to report a bug, but all I hear are features.
Gamma is supposed to be different. The day Gamma becomes the best Power Point with AI isn’t the day it wins, it’s the day it loses.
The workflow is also different. Gamma’s founding team calls it the “Anti-Power Point”, and not because it has AI.
Lemme give you just one example. In the three years I’ve used Gamma, I have never once done the “up, up, up, up, left left left, up, up, up, up, up… oh wait… shift up, up, up, up…” motion to move objects. If I wanted 7 timeline items rather than 8 on a card, I just delete one of them and it’s done. I don’t have to spend 10 minutes moving every object, resize them, shorten the lines, etc. (and then your colleague tells you aw crap we need 9 items after all. It happens every time and you know it).
I spent ⅓ of my life moving PPT objects by inches, and I’m not doing that BS no more.
Could Google Slides or Apple Keynote copy this entire UI/UX? Sure, but they won’t. For one, they wouldn't deign to copy a "small startup".
Second, they literally can't.
Remember that Gamma was designed by literally one of the best design teams in tech. They thought this entire thing through, and decided to impose constraints. There are things you cannot do in Gamma that you can in PPT, because Gamma doesn’t want you to waste your life doing them (like trying every angle of diagonal lines for hours).
What if Google Slides did this? If their users woke up and couldn’t do those things anymore, they’d be pissed. That’s why legacy slideware can’t go there.
It’s like a UI/UX counter-position.
Back to AI Startups
If your startup’s product is just [insert legacy software] with AI, then of course you’d be in deep shite if [insert legacy software] added AI.
But if you’re something different? If you reimagined the way your users performed actions or worked together? If you’ve crafted a digital space that feels like home to them? I suspect you’d be fine.
And then you use AI in the right places to make their work easier.
But do most startups want to tell me about how they've re-imagined work itself? Of course not. They all wanna tell me about how their AI is more powerful and accurate than industry benchmarks, and how their master agent can orchestrate swarms of little agents.
And I get it, I do. The AI hype train is chuchuing along, and the buzzwords can get you some attention.
But I'll just say it one more time.
Building the AI version of an existing software gets you ZERO advantage. Zip.